我们介绍Softmax梯度篡改，一种用于修改神经网络后向通过的梯度的技术，以提高其准确性。我们的方法使用基于功率的概率变换来改变预测的概率值，然后将梯度重新计算在后向通过。这种修改导致更平滑的渐变简介，我们在经验和理论上展示。我们对剩余网络进行了转换参数进行了网格搜索。我们证明修改CUMMNET中的软MAX梯度可能导致培训准确性提高，从而增加训练数据的适合，并最大限度地利用神经网络的学习能力。当与标签平滑等正则化技术相结合时，我们获得更好的测试度量和更低的泛化间隙。 Softmax渐变篡改在ImageNet DataSet上的基线上以0.52 \％$ 0.52 \％$ 0.52 \％$ 0.52 \％。我们的方法非常通用，可以跨各种不同的网络架构和数据集使用。

神经架构的创新促进了语言建模和计算机视觉中的重大突破。不幸的是，如果网络参数未正确初始化，新颖的架构通常会导致挑战超参数选择和培训不稳定。已经提出了许多架构特定的初始化方案，但这些方案并不总是可移植到新体系结构。本文介绍了毕业，一种用于初始化神经网络的自动化和架构不可知论由方法。毕业基础是一个简单的启发式;调整每个网络层的规范，使得具有规定的超参数的SGD或ADAM的单个步骤导致可能的损耗值最小。通过在每个参数块前面引入标量乘数变量，然后使用简单的数字方案优化这些变量来完成此调整。 GradInit加速了许多卷积架构的收敛性和测试性能，无论是否有跳过连接，甚至没有归一化层。它还提高了机器翻译的原始变压器架构的稳定性，使得在广泛的学习速率和动量系数下使用ADAM或SGD来训练它而无需学习速率预热。代码可在https://github.com/zhuchen03/gradinit上获得。

Despite the fact that deep neural networks are powerful models and achieve appealing results on many tasks, they are too large to be deployed on edge devices like smartphones or embedded sensor nodes. There have been efforts to compress these networks, and a popular method is knowledge distillation, where a large (teacher) pre-trained network is used to train a smaller (student) network. However, in this paper, we show that the student network performance degrades when the gap between student and teacher is large. Given a fixed student network, one cannot employ an arbitrarily large teacher, or in other words, a teacher can effectively transfer its knowledge to students up to a certain size, not smaller. To alleviate this shortcoming, we introduce multi-step knowledge distillation, which employs an intermediate-sized network (teacher assistant) to bridge the gap between the student and the teacher. Moreover, we study the effect of teacher assistant size and extend the framework to multi-step distillation. Theoretical analysis and extensive experiments on CIFAR-10,100 and ImageNet datasets and on CNN and ResNet architectures substantiate the effectiveness of our proposed approach.

减少大深度学习模型的处理时间的问题是许多现实世界应用中的根本挑战。早期退出方法通过将附加内部分类器（IC）附加到神经网络的中间层来努力实现这一目标。 IC可以快速返回简单示例的预测，结果，降低整个模型的平均推理时间。但是，如果特定IC不决定早期回答，则其预测被丢弃，其计算有效地浪费。为了解决这个问题，我们引入零时间浪费（ZTW），这是一种新的方法，其中每个IC重用由其前辈返回的预测（1）在IC和（2）之间以相对于类似的方式组合先前输出之间的直接连接。我们对各个数据集和架构进行了广泛的实验，以证明ZTW实现了比最近提出的早期退出方法的其他更好的比例与推理时间权衡。

以前的工作提出了许多新的损失函数和常规程序，可提高图像分类任务的测试准确性。但是，目前尚不清楚这些损失函数是否了解下游任务的更好表示。本文研究了培训目标的选择如何影响卷积神经网络隐藏表示的可转移性，训练在想象中。我们展示了许多目标在Vanilla Softmax交叉熵上导致想象的精度有统计学意义的改进，但由此产生的固定特征提取器转移到下游任务基本较差，并且当网络完全微调时，损失的选择几乎没有效果新任务。使用居中内核对齐来测量网络隐藏表示之间的相似性，我们发现损失函数之间的差异仅在网络的最后几层中都很明显。我们深入了解倒数第二层的陈述，发现不同的目标和近奇计的组合导致大幅不同的类别分离。具有较高类别分离的表示可以在原始任务上获得更高的准确性，但它们的功能对于下游任务不太有用。我们的结果表明，用于原始任务的学习不变功能与传输任务相关的功能之间存在权衡。

近年来，最先进神经网络的参数的数量急剧增加。这种对大规模神经网络感兴趣的激增具有促使新的分布式培训策略的发展，从而实现了这种模型。一种这样的策略是模型平行分布式培训。不幸的是，模型 - 并行性遭受资源利用率差，导致资源浪费。在这项工作中，我们改进了最近的理想化模型 - 并行优化设置：本地学习。由于资源利用率差，我们在当地和全球学习之间介绍了一类中介战略，称为联锁反向化。这些策略保留了本地优化的许多计算效率优势，同时恢复全球优化实现的大部分任务性能。我们评估了我们对图像分类的策略和变压器语言模型，发现我们的策略一致地在任务绩效方面出现本地学习，并在培训效率方面进行全球学习。

Much of the recent progress made in image classification research can be credited to training procedure refinements, such as changes in data augmentations and optimization methods. In the literature, however, most refinements are either briefly mentioned as implementation details or only visible in source code. In this paper, we will examine a collection of such refinements and empirically evaluate their impact on the final model accuracy through ablation study. We will show that, by combining these refinements together, we are able to improve various CNN models significantly. For example, we raise ResNet-50's top-1 validation accuracy from 75.3% to 79.29% on ImageNet. We will also demonstrate that improvement on image classification accuracy leads to better transfer learning performance in other application domains such as object detection and semantic segmentation.

While machine learning is traditionally a resource intensive task, embedded systems, autonomous navigation, and the vision of the Internet of Things fuel the interest in resource-efficient approaches. These approaches aim for a carefully chosen trade-off between performance and resource consumption in terms of computation and energy. The development of such approaches is among the major challenges in current machine learning research and key to ensure a smooth transition of machine learning technology from a scientific environment with virtually unlimited computing resources into everyday's applications. In this article, we provide an overview of the current state of the art of machine learning techniques facilitating these real-world requirements. In particular, we focus on deep neural networks (DNNs), the predominant machine learning models of the past decade. We give a comprehensive overview of the vast literature that can be mainly split into three non-mutually exclusive categories: (i) quantized neural networks, (ii) network pruning, and (iii) structural efficiency. These techniques can be applied during training or as post-processing, and they are widely used to reduce the computational demands in terms of memory footprint, inference speed, and energy efficiency. We also briefly discuss different concepts of embedded hardware for DNNs and their compatibility with machine learning techniques as well as potential for energy and latency reduction. We substantiate our discussion with experiments on well-known benchmark datasets using compression techniques (quantization, pruning) for a set of resource-constrained embedded systems, such as CPUs, GPUs and FPGAs. The obtained results highlight the difficulty of finding good trade-offs between resource efficiency and predictive performance.

自我介绍在训练过程中利用自身的非均匀软监管，并在没有任何运行时成本的情况下提高性能。但是，在训练过程中的开销经常被忽略，但是在巨型模型的时代，培训期间的时间和记忆开销越来越重要。本文提出了一种名为ZIPF标签平滑（ZIPF的LS）的有效自我验证方法，该方法使用网络的直立预测来生成软监管，该软监管在不使用任何对比样本或辅助参数的情况下符合ZIPF分布。我们的想法来自经验观察，即当对网络进行适当训练时，在按样品的大小和平均分类后，应遵循分布的分布，让人联想到ZIPF的自然语言频率统计信息，这是在按样品中的大小和平均值进行排序之后进行的。 。通过在样本级别和整个培训期内强制执行此属性，我们发现预测准确性可以大大提高。使用INAT21细粒分类数据集上的RESNET50，与香草基线相比，我们的技术获得了 +3.61％的准确性增长，而与先前的标签平滑或自我验证策略相比，增益增加了0.88％。该实现可在https://github.com/megvii-research/zipfls上公开获得。

Very deep convolutional networks with hundreds of layers have led to significant reductions in error on competitive benchmarks. Although the unmatched expressiveness of the many layers can be highly desirable at test time, training very deep networks comes with its own set of challenges. The gradients can vanish, the forward flow often diminishes, and the training time can be painfully slow. To address these problems, we propose stochastic depth, a training procedure that enables the seemingly contradictory setup to train short networks and use deep networks at test time. We start with very deep networks but during training, for each mini-batch, randomly drop a subset of layers and bypass them with the identity function. This simple approach complements the recent success of residual networks. It reduces training time substantially and improves the test error significantly on almost all data sets that we used for evaluation. With stochastic depth we can increase the depth of residual networks even beyond 1200 layers and still yield meaningful improvements in test error (4.91% on CIFAR-10).

深度学习使用由其重量进行参数化的神经网络。通常通过调谐重量来直接最小化给定损耗功能来训练神经网络。在本文中，我们建议将权重重新参数转化为网络中各个节点的触发强度的目标。给定一组目标，可以计算使得发射强度最佳地满足这些目标的权重。有人认为，通过我们称之为级联解压缩的过程，使用培训的目标解决爆炸梯度的问题，并使损失功能表面更加光滑，因此导致更容易，培训更快，以及潜在的概括，神经网络。它还允许更容易地学习更深层次和经常性的网络结构。目标对重量的必要转换有额外的计算费用，这是在许多情况下可管理的。在目标空间中学习可以与现有的神经网络优化器相结合，以额外收益。实验结果表明了使用目标空间的速度，以及改进的泛化的示例，用于全连接的网络和卷积网络，以及调用和处理长时间序列的能力，并使用经常性网络进行自然语言处理。

Transferring knowledge from a teacher neural network pretrained on the same or a similar task to a student neural network can significantly improve the performance of the student neural network. Existing knowledge transfer approaches match the activations or the corresponding handcrafted features of the teacher and the student networks. We propose an information-theoretic framework for knowledge transfer which formulates knowledge transfer as maximizing the mutual information between the teacher and the student networks. We compare our method with existing knowledge transfer methods on both knowledge distillation and transfer learning tasks and show that our method consistently outperforms existing methods. We further demonstrate the strength of our method on knowledge transfer across heterogeneous network architectures by transferring knowledge from a convolutional neural network (CNN) to a multi-layer perceptron (MLP) on CIFAR-10. The resulting MLP significantly outperforms the-state-of-the-art methods and it achieves similar performance to the CNN with a single convolutional layer. * Contributed during an internship at Amazon.

分批归一化（BN）由归一化组成部分，然后是仿射转化，并且对于训练深神经网络至关重要。网络中每个BN的标准初始化分别设置了仿射变换量表，并将其转移到1和0。但是，经过训练，我们观察到这些参数从初始化中并没有太大变化。此外，我们注意到归一化过程仍然可以产生过多的值，这对于训练是不可能的。我们重新审视BN公式，并为BN提出了一种新的初始化方法和更新方法，以解决上述问题。实验旨在强调和证明适当的BN规模初始化对性能的积极影响，并使用严格的统计显着性测试进行评估。该方法可以与现有实施方式一起使用，没有额外的计算成本。源代码可在https://github.com/osu-cvl/revisiting-bninit上获得。

Time Series Classification (TSC) is an important and challenging problem in data mining. With the increase of time series data availability, hundreds of TSC algorithms have been proposed. Among these methods, only a few have considered Deep Neural Networks (DNNs) to perform this task. This is surprising as deep learning has seen very successful applications in the last years. DNNs have indeed revolutionized the field of computer vision especially with the advent of novel deeper architectures such as Residual and Convolutional Neural Networks. Apart from images, sequential data such as text and audio can also be processed with DNNs to reach state-of-the-art performance for document classification and speech recognition. In this article, we study the current state-ofthe-art performance of deep learning algorithms for TSC by presenting an empirical study of the most recent DNN architectures for TSC. We give an overview of the most successful deep learning applications in various time series domains under a unified taxonomy of DNNs for TSC. We also provide an open source deep learning framework to the TSC community where we implemented each of the compared approaches and evaluated them on a univariate TSC benchmark (the UCR/UEA archive) and 12 multivariate time series datasets. By training 8,730 deep learning models on 97 time series datasets, we propose the most exhaustive study of DNNs for TSC to date.

模型量化已成为加速深度学习推理的不可或缺的技术。虽然研究人员继续推动量化算法的前沿，但是现有量化工作通常是不可否认的和不可推销的。这是因为研究人员不选择一致的训练管道并忽略硬件部署的要求。在这项工作中，我们提出了模型量化基准（MQBench），首次尝试评估，分析和基准模型量化算法的再现性和部署性。我们为实际部署选择多个不同的平台，包括CPU，GPU，ASIC，DSP，并在统一培训管道下评估广泛的最新量化算法。 MQBENCK就像一个连接算法和硬件的桥梁。我们进行全面的分析，并找到相当大的直观或反向直观的见解。通过对齐训练设置，我们发现现有的算法在传统的学术轨道上具有大致相同的性能。虽然用于硬件可部署量化，但有一个巨大的精度差距，仍然不稳定。令人惊讶的是，没有现有的算法在MQBench中赢得每一项挑战，我们希望这项工作能够激发未来的研究方向。

Although considerable progress has been obtained in neural network quantization for efficient inference, existing methods are not scalable to heterogeneous devices as one dedicated model needs to be trained, transmitted, and stored for one specific hardware setting, incurring considerable costs in model training and maintenance. In this paper, we study a new vertical-layered representation of neural network weights for encapsulating all quantized models into a single one. With this representation, we can theoretically achieve any precision network for on-demand service while only needing to train and maintain one model. To this end, we propose a simple once quantization-aware training (QAT) scheme for obtaining high-performance vertical-layered models. Our design incorporates a cascade downsampling mechanism which allows us to obtain multiple quantized networks from one full precision source model by progressively mapping the higher precision weights to their adjacent lower precision counterparts. Then, with networks of different bit-widths from one source model, multi-objective optimization is employed to train the shared source model weights such that they can be updated simultaneously, considering the performance of all networks. By doing this, the shared weights will be optimized to balance the performance of different quantized models, thus making the weights transferable among different bit widths. Experiments show that the proposed vertical-layered representation and developed once QAT scheme are effective in embodying multiple quantized networks into a single one and allow one-time training, and it delivers comparable performance as that of quantized models tailored to any specific bit-width. Code will be available.

Attention plays a critical role in human visual experience. Furthermore, it has recently been demonstrated that attention can also play an important role in the context of applying artificial neural networks to a variety of tasks from fields such as computer vision and NLP. In this work we show that, by properly defining attention for convolutional neural networks, we can actually use this type of information in order to significantly improve the performance of a student CNN network by forcing it to mimic the attention maps of a powerful teacher network.To that end, we propose several novel methods of transferring attention, showing consistent improvement across a variety of datasets and convolutional neural network architectures. Code and models for our experiments are available at https://github.com/szagoruyko/attention-transfer.

培训深度神经网络是一项非常苛刻的任务，尤其是具有挑战性的是如何适应体系结构以提高训练有素的模型的性能。我们可以发现，有时，浅网络比深网概括得更好，并且增加更多层会导致更高的培训和测试错误。深层残留学习框架通过将跳过连接添加到几个神经网络层来解决此降解问题。最初，需要这种跳过连接才能成功地训练深层网络，因为网络的表达性会随着深度的指数增长而成功。在本文中，我们首先通过神经网络分析信息流。我们介绍和评估批处理循环，该批处理通过神经网络的每一层量化信息流。我们从经验和理论上证明，基于梯度下降的训练方法需要正面批处理融合，以成功地优化给定的损失功能。基于这些见解，我们引入了批处理凝聚正则化，以使基于梯度下降的训练算法能够单独通过每个隐藏层来优化信息流。借助批处理正则化，梯度下降优化器可以将不可吸引的网络转换为可训练的网络。我们从经验上表明，因此我们可以训练“香草”完全连接的网络和卷积神经网络 - 没有跳过连接，批处理标准化，辍学或任何其他建筑调整 - 只需将批处理 - 凝集正则术语添加到500层中损失功能。批处理 - 注入正则化的效果不仅在香草神经网络上评估，还评估了在各种计算机视觉以及自然语言处理任务上的剩余网络，自动编码器以及变压器模型上。

我们通过应用更为理论证明的操作员来寻求改善神经网络中的汇集操作。我们证明Logsumexp提供了用于登录的自然或操作员。当一个人对池中汇集运算符中的元素数正确时，这将成为$ \ text {logavgexp}：= \ log（\ text {mean}（\ exp（x）））$。通过引入单个温度参数，LogavgeXP将其操作数的最大值平滑地过渡到平均值（在限制性情况下发现$ 0 ^ + $和$ t \ to + \ idty $）。在各种深度神经网络架构中，我们在实验测试的LogavgeXP，无论是没有学习的温度参数，都在电脑视觉中的各种深度神经网络架构中。

Batch Normalization (BatchNorm) is a widely adopted technique that enables faster and more stable training of deep neural networks (DNNs). Despite its pervasiveness, the exact reasons for BatchNorm's effectiveness are still poorly understood. The popular belief is that this effectiveness stems from controlling the change of the layers' input distributions during training to reduce the so-called "internal covariate shift". In this work, we demonstrate that such distributional stability of layer inputs has little to do with the success of BatchNorm. Instead, we uncover a more fundamental impact of BatchNorm on the training process: it makes the optimization landscape significantly smoother. This smoothness induces a more predictive and stable behavior of the gradients, allowing for faster training.